Application Of SVM And ANN Fusion Algorithm In Well Logging Lithologic Identificaiton

In the process of petroleum exploration,lithology identification is an important basis for us to search for oil and gas resources.Lithology identification is a process using well logging data to classify the geological rock and soil in oil field into different lithology,helping people understand the formation thickness,structure and lithology change,so as to quickly find the oil-bearing area.Support vector machine is an effective way to deal with well logging data,which is a kind of machine learning method based on statistical learning,good at solving small sample,nonlinear and high dimensional pattern recognition issue.But the numerical accuracy of the core parameter(kernel function and penalty parameter)is not high.In this thesis,support vector machine and artificial neural network fusion algorithm was introduced to build model to recognize and optimize well logging lithology.It has effectively solved the processing and interpretation of logging data and achieved good results.In this thesis,an oil field in China was taken as the research object,a large amount of logging data was selected;dimension-reduced data was obtained through optimization,normalization processing and principal component analysis.There is no need to describe in advance by using artificial neural network.Back propagation can adjust the threshold and weight of the network,and the goal of the smallest network error was finally achieved.The optimal support vector machine penalty parameter and kernel function made it possible for the mix of support vector machine and artificial neural network.The fusion algorithm model was established,and the model was used to identify the lithology of well logging.Recognition experiment of two algorithms were compared between the traditional support vector machine and the genetic support vector machine,and the result showed that support vector machine and artificial neural network fusion algorithm’s well logging lithology recognition rate was higher and its classification ability was stronger,which was a more effective method for logging lithology identification.